CN108931387B - Fault diagnosis method based on multi-sensor signal analysis and capable of providing accurate diagnosis decision - Google Patents
Fault diagnosis method based on multi-sensor signal analysis and capable of providing accurate diagnosis decision Download PDFInfo
- Publication number
- CN108931387B CN108931387B CN201810498072.1A CN201810498072A CN108931387B CN 108931387 B CN108931387 B CN 108931387B CN 201810498072 A CN201810498072 A CN 201810498072A CN 108931387 B CN108931387 B CN 108931387B
- Authority
- CN
- China
- Prior art keywords
- fault
- neural network
- diagnosis
- data
- heat pump
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a fault diagnosis method based on multi-sensor signal analysis, which provides accurate diagnosis decision, and utilizes various sensors to acquire parameters such as pressure, temperature, flow and the like of a heat pump unit in the operation process, and utilizes a vibration sensor to acquire vibration signals of the heat pump unit, so as to comprehensively master the equipment state of an air source heat pump unit. On the basis, various intelligent technical methods are combined, the respective advantages of the intelligent technologies are comprehensively utilized, the advantages are exploited, the disadvantages are avoided, the condition monitoring, the fault diagnosis and the intelligent prediction are carried out on the air source heat pump unit, the sensitivity and the accuracy of a monitoring and diagnosis system can be effectively improved, and the misdiagnosis rate and the missed diagnosis rate are reduced. Meanwhile, a signal processing platform which is convenient to use is designed by adopting a GUI design method based on MATLAB language. Without understanding the system mechanism and analyzing the data, accurate diagnostic decisions are provided for the average operator.
Description
The application is application number: 201510846925.2, filing date: 2015-11-30, divisional application entitled "Fault diagnosis method based on Multi-sensor Signal analysis".
Technical Field
The invention relates to a diagnosis method for acquiring mechanical equipment information by using multiple sensors and analyzing by using a multi-hybrid algorithm, and a high-accuracy and low-misdiagnosis-rate fault diagnosis method is formed by using the diagnosis method.
Background
After the heat pump technology was proposed in 1854 years, the heat pump technology goes through a tortuous development process, and has now entered a comprehensive high-speed development stage, particularly under the environmental pressure of energy crisis and global warming, the heat pump technology becomes a focus of attention of all countries, and research, application and popularization of various heat pump technologies have also risen to a height of continuous attention. As an energy-saving and emission-reducing technology, the heat pump has a wide technical prospect, more and more countries, governments and enterprises can realize energy-saving and environment-friendly benefits brought by the heat pump, and market data also shows that the future development trend is good. The research on the application of the heat pump technology in the actual engineering can better play the energy-saving and environment-friendly functions of the heat pump in practice, and is an important basis for popularizing the heat pump technology, so that the actual engineering application and the optimization research work are very important.
In recent years, heat pump hot water systems are widely applied in China to meet the requirements of users for domestic hot water, but various enemy faults can occur in the operation of the heat pump hot water systems inevitably due to aging of equipment in the heat pump hot water systems or other reasons. At present, system fault elimination and equipment maintenance mainly depend on the experience of field personnel and relevant instruments to complete fault diagnosis, and whether faults can be eliminated in time is restricted by the level and experience of the field personnel. And the method is only limited to local diagnosis of a fault part, lacks systematic diagnosis maintenance and fault prediction methods, and fault reason searching and troubleshooting account for more than 50% of total fault elimination time according to experience. Therefore, the establishment of a perfect fault diagnosis system has important significance for the rapid and effective intelligent diagnosis of the ground source heat pump hot water system.
The fault diagnosis is essentially a pattern recognition problem and comprises three links of signal acquisition, feature extraction and selection and state recognition, wherein the signal acquisition is the premise of fault diagnosis, the feature extraction and selection are the key points of fault diagnosis, and the state recognition is the core of diagnosis. The accuracy of fault diagnosis can be improved only if the correctness of the acquired signals, the sensitivity of the extracted and selected features and the effectiveness of the state identification method are ensured.
Some parameters of the machine that reflect possible faults of the machine, such as temperature, pressure, flow, etc., may be obtained using measurements of the sensors. The vibration and noise and its characteristic information generated by the mechanical equipment in the course of operation are the main signals reflecting the change of mechanical equipment and its operation state. One or more signals may be obtained by using sensors to learn the state of machine operation. The acquisition of various sensor signals can help to comprehensively grasp the mechanical operation state, so that the equipment state of the air source heat pump unit can be comprehensively mastered.
At present, the fault diagnosis methods for heat pump units are few, or only a single fault diagnosis method is adopted, and system fault judgment is carried out according to data acquired by certain single sensors, so that the result is not only lack of accuracy, but also misdiagnosis exists. Therefore, various sensor signals of the same system at a certain moment are obtained, and various intelligent diagnosis technologies are utilized to make up for the deficiencies, so that the problems of the single intelligent diagnosis technology can be effectively solved, the accuracy and the sensitivity of the diagnosis and monitoring system are improved, and the misdiagnosis rate is reduced.
The state identification method based on a single intelligent classifier is difficult to accurately diagnose early faults and compound faults of complex mechanical equipment. In order to overcome the defects of a single intelligent classifier and improve the diagnosis confirming rate of the fault diagnosis of the complex equipment, a hybrid intelligent diagnosis method combining a plurality of intelligent classifiers is required. Multi-classifier combination diagnostics is one of the modes to implement hybrid intelligent diagnostics. Different sets of input features can be obtained by using different pre-processing techniques or feature extraction methods, and several classifiers that input different sets of features often exhibit complementary classification characteristics. Therefore, if several classifier output results of different input feature sets are synthesized by an integration technique, the final result can be better than the best single classifier.
The knowledge of experts in the field of heat pump hot water is software-based, and the artificial intelligence algorithm is adopted to identify faults, so that the normal operation of the control system can be better ensured. The fault diagnosis method based on knowledge in the artificial intelligence diagnosis technology is widely applied due to the characteristics of intelligence and independence on mathematical models.
Jiangqiang et al, Harbin university, studied the fault diagnosis of a neural network-based air source heat pump unit, applied a BP neural network model to simulate the performance of the heat pump unit, and trained the neural network using symptom examples from simulation experiments and knowledge in the expert field.
The artificial neural network method is applied by Zhang Zhonghe, Wangkang and other people of the science and technology university of Xian architecture to diagnose the fault of the underground water source heat pump system. The difference compared to previous studies is that they place the center of gravity of the study on "soft failures" that have an impact on energy efficiency and are not easily perceived. The method has positive significance for the operation and maintenance of the underground water source heat pump system.
Summarizing the research results at home and abroad, the application of the fault diagnosis technology in the field of heat pump units in China has been developed and applied to a certain extent, but the following problems still exist:
most of the existing hybrid intelligent diagnosis research still uses discrete physical signals such as temperature, pressure and the like to form diagnosis information, rarely uses continuous signals such as vibration information or sound information and the like, and causes the imperfection of the diagnosis information. Mechanical system faults are often expressed in dynamics, acoustics, tribology, thermodynamics and other multi-physical fields, so that diagnosis information is incomplete only by utilizing a certain physical field, and missed detection and misjudgment of the faults are inevitable.
From the point of view of diagnostic methods, single fault diagnostic methods have inevitable drawbacks as research and application progresses. The application of the single intelligent technology is limited by the problems of incomplete diagnosis information, artificial determination of fuzzy membership function, difficulty in acquiring knowledge of an expert system, lack of fault sample training of a neural network and the like. Therefore, the combination of multiple diagnostic methods can make the knowledge expression in the diagnostic system more perfect and clear, and the accuracy of diagnosis can be greatly improved and the diagnosis can be performed regardless of a specific diagnostic method.
Disclosure of Invention
The invention aims to provide a fault diagnosis method based on multi-sensor signal analysis, which can effectively diagnose and predict the fault mode of a complex machine.
The technical solution of the invention is as follows:
a fault diagnosis method based on multi-sensor signal analysis is characterized in that: comprises the following steps:
(1) the single chip microcomputer is used for driving a plurality of sensors to collect unit data, unit operation parameters comprise water supply temperature, water return temperature, water temperature of a water tank and water level of the water tank, serial port communication is carried out between the unit operation parameters and the PC, and the collected data are sent to the PC;
(2) the method comprises the following steps of (1) acquiring a vibration signal of a unit, wherein a wireless sensor is adopted, and vibration data are transmitted to a computer through a wireless module;
(3) acquiring pressure, temperature and flow parameters of the heat pump unit measured by each sensor in different transition states in the operation process, and performing fusion operation to form characteristic vectors in different states;
(4) carrying out mode mapping on a feature vector and a state formed by acquired sensor data of a plurality of moments or a plurality of running states to obtain a relation between a fault and a symptom, and designing and training a neural network to classify the fault so that a mode mapping process can be completed according to different fault symptoms when a system carries out fault diagnosis;
(5) determining the structure of a BP neural network optimized by a genetic algorithm according to the feature vector dimension of data fusion, completing the training process of the neural network, and performing related optimization according to the training result of the neural network;
(6) optimizing the initial weight and threshold distribution of the BP neural network by adopting a genetic algorithm, and finding the optimal weight and threshold of the BP neural network through selection, intersection and variation operations; training the optimized BP neural network by using test data to obtain a final optimized BP neural network fault classifier;
(7) carrying out denoising related operation on vibration signals obtained in a corresponding mechanical state, and then carrying out wavelet decomposition on vibration waveforms by adopting a wavelet packet analysis method to obtain reconstructed wavelet coefficients and reconstructed wavelet energy spectrums of all nodes after three layers of wavelet decomposition;
(8) extracting energy, variance and wavelet coefficient data of each node from the obtained reconstructed wavelet coefficient and the reconstructed wavelet energy spectrum, and fusing the data to obtain a characteristic vector of the vibration signal;
(9) carrying out feature reduction on the feature vector of the vibration signal by using a rough set theory, and carrying out condition attribute reduction and decision rule reduction on a decision table to obtain a minimum decision rule as a final classification rule;
(10) designing and training a counter propagation neural network according to the rules after rough set reduction, and acquiring a fault mode classifier of a vibration signal after the training is successful;
(11) entering a fault decision fusion stage, taking the feature vector of the signal obtained in S3 as input, obtaining a fault result by using the neural network fault classifier trained in the step (6), and diagnosing by using the fault mode classifier of the vibration signal obtained in the step (10) according to the feature vector of the vibration signal obtained in the step (8) to obtain a diagnosis result; finally, performing D-S decision fusion on the diagnosis results of the two different signals, wherein the final fusion result is the diagnosis result;
(12) finally, designing a fault diagnosis signal processing platform by adopting a GUI design method based on an MATLAB language; technicians can obtain the current unit operation parameters through simple operation, make fault judgment in time, save the current data and check historical data.
The invention has the beneficial effects that:
1. the method comprises the steps of utilizing various sensors to acquire parameters such as pressure, temperature and flow of the heat pump unit in the operation process, and utilizing a vibration sensor to acquire vibration signals of the heat pump unit, so as to comprehensively master the equipment state of the air source heat pump unit.
2. The intelligent monitoring system combines multiple intelligent technical methods, comprehensively utilizes the respective advantages of the intelligent technologies, increases the length and avoids the shortages, carries out state monitoring, fault diagnosis and intelligent prediction on the air source heat pump unit, can effectively improve the sensitivity and the accuracy of the monitoring and diagnosis system, and reduces the misdiagnosis rate and the missed diagnosis rate.
3. Based on MATLAB language, a GUI design method is adopted to design a signal processing platform which is convenient to use. Without understanding the system mechanism and analyzing the data, accurate diagnostic decisions are provided for the average operator.
Drawings
The invention is further illustrated by the following figures and examples.
FIG. 1 is a schematic diagram of a fault diagnosis system based on multi-sensor signal analysis of a heat pump unit according to the present invention.
FIG. 2 is a multi-sensor fault diagnostic model.
Fig. 3 is a structural diagram of a BP neural network designed in the present invention.
FIG. 4 is a block diagram of a coarse set attribute reduction algorithm flow.
FIG. 5 is a main interface of the heat pump unit fault diagnosis system.
Various reference numbers in the above figures: p1-high pressure, P2-low pressure, P3-condensing temperature, P4-evaporating temperature, P5-suction superheat temperature, P6-liquid subcooling temperature, P7-discharge temperature and P8-water flow temperature difference through the condenser. T1-no fault, T2-refrigerant leak, T3-compressor discharge valve leak, T4-liquid line blocked, T5-condenser fouling, T6-evaporator fouling.
Detailed Description
Firstly, parameters of the unit running at a certain moment, including temperature, pressure and other data, are fused into a characteristic vector at the moment, data of a plurality of sensors at a plurality of moments or a plurality of mechanical states are obtained, the relation between faults and symptoms is obtained, and a mode mapping process is completed according to different fault symptoms. The number of input layer nodes and the number of output layer nodes of the BP neural network are determined by taking the BP neural network as a training set of the BP neural network, and the number of the input layer nodes and the number of the output layer nodes of the BP neural network are determined according to an empirical formula
The number of hidden layer nodes is determined. Common soft faults of the air source heat pump unit are: refrigerant leakage, compressor discharge valve leakage, blocked liquid lines, condenser fouling, and evaporator fouling. First, input sample data and target output of the network are determined. There are 8 characteristic quantities for a heat pump unit: high pressure P1, low pressure P2, condensation temperature P3, evaporation temperatureP4, suction superheat temperature P5, liquid subcooling temperature P6, discharge temperature P7 and water flow differential across the condenser P8. Data obtained by simulation tests are used as input, and output is no-fault T1, refrigerant leakage T2, compressor vent valve leakage T3, liquid line blockage T4, condenser fouling T5 and evaporator fouling T6. The network structure is then determined. And adopting a BP neural network to carry out fault diagnosis. Designing the number of nodes of an input layer of the BP network to be 8, the number of nodes of an output layer to be 6, and determining the number of nodes of a hidden layer to be 13. And determining the structure of the BP neural network to be 8-13-6. The structure diagram is shown in figure 3.
But because the BP network tends to get into local minima, the desired training result is not achieved during the training process. Neural networks have their own drawbacks and deficiencies, which need to be optimized. Because the genetic algorithm has global search capability and stronger robustness, the network training can be enhanced, the network performance can be improved, and the global optimal solution can be found more easily. Therefore, genetic algorithms are used to optimize the initial values and thresholds of the BP neural network. The initial connection weight and the threshold of each layer before the training of the BP neural network are random values between [0,1], and the unoptimized random values can slow down the convergence rate of the BP neural network and easily make the final result be a non-optimal solution. The basic idea of genetic algorithm optimization BP neural network is as follows: and optimizing the initial weight and threshold distribution of the BP neural network by adopting a genetic algorithm, and finding the optimal weight and threshold of the BP neural network through selection, intersection and variation operations. And training the BP neural network by using the new weight and the threshold value to obtain a better training result.
And simultaneously, processing the vibration signals in the plurality of mechanical states. The method comprises the steps of collecting vibration signals by using a vibration sensor in a heat pump unit no-fault T1 and fault T2-T6 mode, decomposing the vibration signals by using a wavelet packet analysis algorithm, and extracting characteristic vectors of all waveforms. Wavelet packet decomposition follows the law of conservation of energy, and the signal in each band represents the information of the original signal in that frequency band. The wavelet packet has higher resolution, the energy value of a specific frequency band is obtained after deep decomposition of the signal is realized, and the obtained characteristic value is utilized to form a diagnostic characteristic vector. Wavelet packet decomposition of signals can be realized by using various wavelet packet bases, and an optimal wavelet packet base can be obtained by spending a cost function. The signal can be characterized by wavelet packet coefficients, and the size of the coefficients can represent the contribution of information in the wavelet packet basis to the signal. The best wavelet packet basis is obtained by using Shannon entropy criterion as cost function.
Some redundant features may exist in the original features extracted by wavelet packet decomposition, and the redundant features may make the fault classification in the later period more complicated and have an influence on the classification accuracy. Therefore, it is necessary to reduce and select the extracted features, and to select sensitive features as input to the classifier without losing failure information. The reduction is to obtain a minimum decision rule as a final classification rule by carrying out condition attribute reduction and decision rule reduction on the decision table by using a rough set theory. The rough set attribute reduction algorithm flow diagram is shown in fig. 4. And taking the reduced feature vectors and the corresponding fault modes as a training set of the classifier, and determining the structure of the classifier.
Entering a fault decision fusion stage, and fusing by using a Dempster-Shafer (D-S for short) evidence theory fusion algorithm. The core idea of decision-level fusion is that each sensor makes a local decision according to the respective observation result, then the decision result is transmitted to a fusion center through a communication channel, and then the fusion center performs comprehensive processing on each local decision to make a final decision. Two sets of evidence E under the assumed recognition framework omega1And E2The corresponding basic trust distribution function is m1And m2The jiao Yuan is AiAnd BjThe D-S synthesis rule is as follows:
several independent pieces of evidence can be combined using the D-S synthesis rule. Compared with single set independent processing, the information fusion can improve the reliability of decision results, reduce inference fuzzy degree, improve detection precision, improve spatial resolution, and enhance fault tolerance and adaptivity of the system, thereby improving the performance of the whole system. The final fusion result is the diagnosis result.
And finally, designing a fault diagnosis signal processing platform by adopting a GUI design method based on an MATLAB language. Technicians can obtain current unit operation parameters through simple operation, timely make fault judgment, save current data and check historical data. The main interface is shown in figure 5.
The invention takes a certain air source heat pump unit as a research object, utilizes various sensors to collect parameters such as pressure, temperature, flow and the like of the heat pump unit in the running process, and utilizes a vibration sensor to collect vibration signals of the unit, thereby comprehensively mastering the equipment state of the air source heat pump unit. On the basis, a plurality of intelligent technical methods are combined to carry out state monitoring, fault diagnosis and intelligent prediction on the air source heat pump unit. Meanwhile, a signal processing platform convenient to use is designed by adopting a GUI design method based on MATLAB language.
The fault diagnosis method based on multi-sensor signal analysis comprises the following steps:
(1) the single chip microcomputer is used for driving the plurality of sensors to collect unit data, unit operation parameters comprise water supply temperature, water return temperature, water tank water level and the like, serial port communication is carried out between the unit operation parameters and the PC, and the collected data are sent to the PC.
(2) The collection of the vibration signals of the unit can adopt a wireless sensor, and the power consumption is low. And transmitting the vibration data to a computer through a wireless module.
(S): and acquiring parameters such as pressure, temperature, flow and the like of the heat pump unit measured by each sensor in different transition states in the operation process, and performing fusion operation to form characteristic vectors in different states.
(4) And carrying out mode mapping on the characteristic vectors and the states formed by the acquired sensor data of a plurality of moments or a plurality of running states to obtain the relation between faults and symptoms, and designing and training a neural network to classify the faults so that the mode mapping process can be completed according to different fault symptoms when the system carries out fault diagnosis.
(5) And determining the structure of the BP neural network optimized by the genetic algorithm according to the feature vector dimension of the data fusion, and finishing the training process of the neural network. But because the BP network tends to trap into the sector very small, the desired training result is not achieved during the training process. Therefore, the relevant optimization is performed according to the training result of the neural network.
(6) And optimizing the initial weight and threshold distribution of the BP neural network by adopting a genetic algorithm, and finding the optimal weight and threshold of the BP neural network through selection, intersection and variation operations. The optimized BP neural network is trained using test data. And obtaining the finally optimized BP neural network fault classifier.
(7) After relevant operations such as denoising and the like are carried out on the vibration signals obtained in the corresponding mechanical state, wavelet packet analysis is adopted to carry out wavelet decomposition on the vibration waveforms, and the reconstructed wavelet coefficients and the reconstructed wavelet energy spectrums of all nodes after three-layer wavelet decomposition are obtained.
(8) And extracting data such as energy, variance, wavelet coefficient and the like of each node from the acquired reconstruction wavelet coefficient and reconstruction wavelet energy spectrum, and fusing the data to obtain the characteristic vector of the vibration signal.
(9) There may be some redundant features in the original features extracted by wavelet packet decomposition, and the redundant features make the later fault classification more complicated, so it is necessary to reduce and select the extracted features. The feature vectors of the vibration signals can be subjected to feature reduction by using a rough set theory, and the minimum decision rule is obtained by performing condition attribute reduction and decision rule reduction on the decision table and is used as a final classification rule.
(10) Designing and training the counter propagation neural network according to the rules after rough set reduction, and successfully training. And a fault mode classifier for obtaining the vibration signal.
(11) And (4) entering a fault decision fusion stage, taking the feature vector of the signal obtained in the step S3 as input, obtaining a fault result by using the neural network fault classifier trained in the step S6, and diagnosing the feature vector of the vibration signal obtained in the step (10) by using the fault mode classifier of the vibration signal obtained in the step S8 to obtain a diagnosis result. And finally, performing D-S decision fusion on the diagnosis results of the two different signals, wherein the final fusion result is the diagnosis result.
(12) And finally, designing a fault diagnosis signal processing platform by adopting a GUI design method based on an MATLAB language. Technical staff can obtain the current unit operation parameters through simple operation, make fault judgment in time, save the current data, check historical data, provide more reliable fault sign information for the subsequent fault diagnosis, and can judge the mechanical state of the unit more quickly and accurately.
It should be appreciated by those skilled in the art that the above embodiments are only for illustrating the technical solutions of the present invention, and not for limiting the present invention, and any changes and modifications to the above embodiments based on the spirit of the present invention will fall within the protection scope of the claims of the present invention.
Claims (1)
1. A fault diagnosis method based on multi-sensor signal analysis and capable of providing accurate diagnosis decision is characterized in that:
comprises the following steps:
(1) the single chip microcomputer is used for driving a plurality of sensors to collect unit data, unit operation parameters comprise water supply temperature, water return temperature, water temperature of a water tank and water level of the water tank, serial port communication is carried out between the unit operation parameters and the PC, and the collected data are sent to the PC;
(2) the method comprises the following steps of (1) acquiring a vibration signal of a unit, wherein a wireless sensor is adopted, and vibration data are transmitted to a computer through a wireless module;
(3) acquiring pressure, temperature and flow parameters of the heat pump unit measured by each sensor in different conversion states in the operation process, and performing fusion operation to form characteristic vectors in different states;
(4) carrying out mode mapping on a feature vector and a state formed by acquired sensor data of a plurality of moments or a plurality of running states to obtain a relation between a fault and a symptom, and designing and training a neural network to classify the fault so that a mode mapping process can be completed according to different fault symptoms when a system carries out fault diagnosis;
(5) determining the structure of a BP neural network optimized by a genetic algorithm according to the feature vector dimension of data fusion, completing the training process of the neural network, and performing related optimization according to the training result of the neural network;
(6) optimizing the initial weight and threshold distribution of the BP neural network by adopting a genetic algorithm, and finding the optimal weight and threshold of the BP neural network through selection, intersection and variation operations; training the optimized BP neural network by using test data to obtain a final optimized BP neural network fault classifier;
(7) carrying out denoising related operation on vibration signals obtained in a corresponding mechanical state, and then carrying out wavelet decomposition on vibration waveforms by adopting a wavelet packet analysis method to obtain a reconstructed wavelet coefficient and a reconstructed wavelet energy spectrum of each node after three-layer wavelet decomposition;
(8) extracting energy, variance and wavelet coefficient data of each node from the obtained reconstructed wavelet coefficient and the reconstructed wavelet energy spectrum, and fusing the data to obtain a characteristic vector of the vibration signal;
(9) carrying out feature reduction on the feature vector of the vibration signal by using a rough set theory, and carrying out condition attribute reduction and decision rule reduction on a decision table to obtain a minimum decision rule as a final classification rule;
(10) designing and training a counter propagation neural network according to the rules after rough set reduction, and acquiring a fault mode classifier of a vibration signal after the training is successful;
(11) entering a fault decision fusion stage, taking the feature vector of the signal obtained in S3 as input, obtaining a fault result by using the neural network fault classifier trained in the step (6), and diagnosing by using the fault mode classifier of the vibration signal obtained in the step (10) according to the feature vector of the vibration signal obtained in the step (8) to obtain a diagnosis result; finally, performing D-S decision fusion on the diagnosis results of the two different signals, wherein the final fusion result is the diagnosis result;
(12) finally, designing a fault diagnosis signal processing platform by adopting a GUI design method based on an MATLAB language; technicians can obtain current unit operation parameters through simple operation, make fault judgment in time, save current data and check historical data;
wherein the specific steps of the steps (1) - (5) are as follows:
firstly, fusing data such as temperature and pressure of a certain time operation parameter of a unit into a characteristic vector of the time, acquiring data of a plurality of sensors at a plurality of times or a plurality of mechanical states, obtaining a relation between a fault and a symptom, and completing a mode mapping process according to different fault symptoms; the number of nodes of an input layer and the number of nodes of an output layer of the BP neural network are determined by taking the BP neural network as a training set of the BP neural network, and the number of the nodes of the input layer and the number of the nodes of the output layer of the BP neural network are determined according to an empirical formula
Determining the number of hidden layer nodes; common soft faults of the air source heat pump unit are: refrigerant leakage, compressor vent valve leakage, blocked liquid line, condenser fouling and evaporator fouling; firstly, input sample data and target output of a network are determined; there are 8 characteristic quantities for a heat pump unit: high pressure P1, low pressure P2, condensation temperature P3, evaporation temperature P4, suction superheat temperature P5, liquid subcooling temperature P6, discharge temperature P7 and water flow temperature difference P8 through the condenser; data obtained through simulation tests are used as input, and output is no-fault T1, refrigerant leakage T2, compressor exhaust valve leakage T3, liquid pipeline blockage T4, condenser fouling T5 and evaporator fouling T6; then determining a network structure, adopting a BP neural network to carry out fault diagnosis, designing the number of input layer nodes of the BP network to be 8, the number of output layer nodes to be 6, determining the number of hidden layer nodes to be 13, and determining the structure of the BP neural network to be 8-13-6;
the specific steps of the step (11) are as follows:
entering a fault decision fusion stage, and fusing by using a Dempster-Shafer evidence theory fusion algorithm; the core idea of decision-level fusion is as follows: each sensor makes a local judgment according to the respective observation result and then transmits the judgment result through a communication channelSending the judgment result to a fusion center, and performing comprehensive treatment on each local judgment by the fusion center to make a final judgment; two sets of evidence E under the assumed recognition framework omega1And E2The corresponding basic trust distribution function is m1And m2The jiao Yuan is AiAnd BjThe D-S synthesis rule is as follows:
several independent evidences can be combined using the D-S synthesis rule; the final fusion result is the diagnosis result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810498072.1A CN108931387B (en) | 2015-11-30 | 2015-11-30 | Fault diagnosis method based on multi-sensor signal analysis and capable of providing accurate diagnosis decision |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510846925.2A CN105372087B (en) | 2015-11-30 | 2015-11-30 | Method for diagnosing faults based on multiple sensor signals analysis |
| CN201810498072.1A CN108931387B (en) | 2015-11-30 | 2015-11-30 | Fault diagnosis method based on multi-sensor signal analysis and capable of providing accurate diagnosis decision |
Related Parent Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510846925.2A Division CN105372087B (en) | 2015-11-30 | 2015-11-30 | Method for diagnosing faults based on multiple sensor signals analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108931387A CN108931387A (en) | 2018-12-04 |
| CN108931387B true CN108931387B (en) | 2020-05-12 |
Family
ID=55374485
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810498072.1A Active CN108931387B (en) | 2015-11-30 | 2015-11-30 | Fault diagnosis method based on multi-sensor signal analysis and capable of providing accurate diagnosis decision |
| CN201510846925.2A Active CN105372087B (en) | 2015-11-30 | 2015-11-30 | Method for diagnosing faults based on multiple sensor signals analysis |
Family Applications After (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510846925.2A Active CN105372087B (en) | 2015-11-30 | 2015-11-30 | Method for diagnosing faults based on multiple sensor signals analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN108931387B (en) |
Families Citing this family (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105915891B9 (en) * | 2016-05-06 | 2018-08-24 | 哈尔滨工程大学 | A kind of imaging sensor key parameter test method based on details in a play not acted out on stage, but told through dialogues variance signal |
| CN106369834A (en) * | 2016-09-02 | 2017-02-01 | 南通大学 | Directly-heated type heat pump system constant-temperature flow control method based on neural network |
| CN106482938B (en) * | 2016-10-14 | 2018-08-03 | 温州大学 | Brake fluid system multi-source fusion failure predicting method based on GA-BP networks |
| CN107606739B (en) * | 2017-08-15 | 2019-07-26 | 清华大学 | The fault detection method of heating ventilation air-conditioning system |
| CN107730000A (en) | 2017-09-15 | 2018-02-23 | 珠海格力电器股份有限公司 | Refrigerant leakage detection method and device for air conditioner |
| CN107830997A (en) * | 2017-10-25 | 2018-03-23 | 华北电力大学（保定） | Equipment fault diagnosis method based on industrial wireless sensor network and wavelet transformation |
| US11947863B2 (en) | 2018-02-28 | 2024-04-02 | Robert Bosch Gmbh | Intelligent audio analytic apparatus (IAAA) and method for space system |
| CN108958221A (en) * | 2018-07-12 | 2018-12-07 | 华北电力大学（保定） | Equipment fault diagnosis method based on industrial Internet of Things Yu support vector machines multi-classification algorithm |
| CN109297077B (en) * | 2018-09-03 | 2020-04-21 | 西安石油大学 | Method for monitoring step utilization system of middle-deep layer non-interference ground rock heat supply system |
| CN109764476B (en) * | 2018-12-18 | 2020-09-04 | 珠海格力电器股份有限公司 | Refrigerant leakage monitoring method and system and air conditioning unit |
| CN109858352B (en) * | 2018-12-26 | 2020-09-18 | 华中科技大学 | Fault diagnosis method based on compressed sensing and improved multi-scale network |
| CN110008914A (en) * | 2019-04-11 | 2019-07-12 | 杨勇 | A kind of pattern recognition system neural network based and recognition methods |
| CN110108509A (en) * | 2019-04-28 | 2019-08-09 | 西安建筑科技大学 | A kind of sewage source heat pump unit intelligent failure diagnosis method |
| CN110110424A (en) * | 2019-04-29 | 2019-08-09 | 西南石油大学 | A kind of compressor adaptive performance curve generation method |
| CN110135081A (en) * | 2019-05-20 | 2019-08-16 | 哈尔滨工程大学 | A kind of internal combustion engine tribology Expert System Design method based on artificial intelligence |
| CN110210459A (en) * | 2019-06-24 | 2019-09-06 | 北京航空航天大学 | A kind of prediction technique and prediction meanss of engine valve clearance |
| CN111307453B (en) * | 2020-03-20 | 2021-11-12 | 朗斯顿科技(北京)有限公司 | Transmission system fault diagnosis method based on multi-information fusion |
| CN111297178B (en) * | 2020-04-17 | 2022-03-29 | 青岛海尔智慧厨房电器有限公司 | Steam box and control method thereof |
| CN111967486A (en) * | 2020-06-02 | 2020-11-20 | 安徽三禾一信息科技有限公司 | Complex equipment fault diagnosis method based on multi-sensor fusion |
| CN112598019B (en) * | 2020-10-28 | 2024-04-02 | 南京航空航天大学 | Heterogeneous vibration signal space-time multi-cross-correlation analysis and self-adaptive weighted fusion method |
| CN112894882A (en) * | 2020-12-30 | 2021-06-04 | 哈尔滨工业大学芜湖机器人产业技术研究院 | Robot fault detection system based on industrial internet |
| CN112733692B (en) * | 2021-01-04 | 2021-11-30 | 润联智慧科技(西安)有限公司 | Fault prediction method and device based on integrated hybrid model and related equipment |
| CN112748317B (en) * | 2021-03-23 | 2022-03-25 | 国网河南省电力公司电力科学研究院 | Switch cabinet partial discharge fault detection method and system based on multiple monitoring data |
| CN113359623B (en) * | 2021-05-08 | 2022-08-23 | 深圳有象智联科技有限公司 | Method and device for monitoring working state and computer readable storage medium |
| CN113188235A (en) * | 2021-05-28 | 2021-07-30 | 西安建筑科技大学 | Multi-source fault diagnosis method, system and equipment for sensor in air conditioning system |
| CN113640027B (en) * | 2021-08-12 | 2024-05-07 | 常州英集动力科技有限公司 | Heat exchange unit fault diagnosis method and system based on edge calculation and neural network |
| CN114882680A (en) * | 2022-04-19 | 2022-08-09 | 深圳闪回科技有限公司 | Remote terminal management system and method |
| CN115421073B (en) * | 2022-08-29 | 2024-08-16 | 吉林大学 | Identification method for pseudo-data of normalized curve |
| CN115374881B (en) * | 2022-10-17 | 2023-01-17 | 图林科技(深圳)有限公司 | Fault diagnosis method for running state of refrigeration equipment |
| CN116538092B (en) * | 2023-07-06 | 2023-11-14 | 中国科学院理化技术研究所 | Compressor on-line monitoring and diagnosing method, device, equipment and storage medium |
| CN117780679B (en) * | 2024-02-26 | 2024-05-03 | 无锡精恩风机有限公司 | Centrifugal fan fault diagnosis method and online monitoring system |
| CN118152730A (en) * | 2024-03-11 | 2024-06-07 | 太极计算机股份有限公司 | Hydraulic electromechanical equipment fault diagnosis method and device, electronic equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101867486A (en) * | 2010-06-08 | 2010-10-20 | 江苏大学 | Wireless sensor network fault diagnosis method |
| CN104330255A (en) * | 2014-11-04 | 2015-02-04 | 徐州隆安光电科技有限公司 | Gear fault diagnosing method based on multi-sensor information fusion |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1166922C (en) * | 2002-07-18 | 2004-09-15 | 上海交通大学 | Multiple-sensor and multiple-object information fusing method |
| CN102736562B (en) * | 2012-07-10 | 2014-06-04 | 北京信息科技大学 | Knowledge base construction method oriented to fault diagnosis and fault prediction of numerical control machine tool |
| CN102944416B (en) * | 2012-12-06 | 2015-04-01 | 南京匹瑞电气科技有限公司 | Multi-sensor signal fusion technology-based fault diagnosis method for wind turbine blades |
| JP5530020B1 (en) * | 2013-11-01 | 2014-06-25 | 株式会社日立パワーソリューションズ | Abnormality diagnosis system and abnormality diagnosis method |
| CN103971162A (en) * | 2014-04-04 | 2014-08-06 | 华南理工大学 | Method for improving BP (back propagation) neutral network and based on genetic algorithm |
| CN103900816A (en) * | 2014-04-14 | 2014-07-02 | 上海电机学院 | Method for diagnosing bearing breakdown of wind generating set |
| CN104503235B (en) * | 2014-12-09 | 2017-05-10 | 中国石油大学(华东) | Condition monitoring method based on improved BP neural network for power plant equipment |
| CN104636493A (en) * | 2015-03-04 | 2015-05-20 | 浪潮电子信息产业股份有限公司 | Dynamic data grading method based on multi-classifier fusion |
| CN104833534A (en) * | 2015-04-21 | 2015-08-12 | 广州市地下铁道总公司 | Train running fault diagnosis device based on multi-source information fusion, and method |
| CN104834828A (en) * | 2015-05-26 | 2015-08-12 | 重庆大学 | Method for diagnosing physiological abnormality of old people based on DS evidence theory-neural network algorithm |
| CN104866904B (en) * | 2015-06-16 | 2019-01-01 | 中电科软件信息服务有限公司 | A kind of BP neural network parallel method of the genetic algorithm optimization based on spark |
-
2015
- 2015-11-30 CN CN201810498072.1A patent/CN108931387B/en active Active
- 2015-11-30 CN CN201510846925.2A patent/CN105372087B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101867486A (en) * | 2010-06-08 | 2010-10-20 | 江苏大学 | Wireless sensor network fault diagnosis method |
| CN104330255A (en) * | 2014-11-04 | 2015-02-04 | 徐州隆安光电科技有限公司 | Gear fault diagnosing method based on multi-sensor information fusion |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105372087A (en) | 2016-03-02 |
| CN105372087B (en) | 2018-06-22 |
| CN108931387A (en) | 2018-12-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108931387B (en) | Fault diagnosis method based on multi-sensor signal analysis and capable of providing accurate diagnosis decision | |
| CN110555479B (en) | Fault feature learning and classifying method based on 1DCNN and GRU fusion | |
| CN107941537B (en) | A kind of mechanical equipment health state evaluation method | |
| CN110262463B (en) | Rail transit platform door fault diagnosis system based on deep learning | |
| CN101539241B (en) | Hierarchical multi-source data fusion method for pipeline linkage monitoring network | |
| CN110598851A (en) | Time series data abnormity detection method fusing LSTM and GAN | |
| CN109827777A (en) | Rolling bearing fault prediction technique based on Partial Least Squares extreme learning machine | |
| CN112016251B (en) | Nuclear power device fault diagnosis method and system | |
| Chen et al. | Deep attention relation network: A zero-shot learning method for bearing fault diagnosis under unknown domains | |
| CN109213127A (en) | A kind of HVAC system gradual failure diagnostic method based on deep learning | |
| CN109165604A (en) | The recognition methods of non-intrusion type load and its test macro based on coorinated training | |
| Zhang et al. | Imbalanced data based fault diagnosis of the chiller via integrating a new resampling technique with an improved ensemble extreme learning machine | |
| CN115187832A (en) | Energy system fault diagnosis method based on deep learning and gram angular field image | |
| CN103953490A (en) | Implementation method for monitoring status of hydraulic turbine set based on HLSNE | |
| CN103323228A (en) | Mining drill fault intelligent identification method | |
| CN109389325A (en) | Transformer substation electronic transducer state evaluating method based on wavelet neural network | |
| CN115791174B (en) | Rolling bearing abnormality diagnosis method, system, electronic equipment and storage medium | |
| CN108897936A (en) | A kind of sewage source heat pump unit method for diagnosing faults based on PSO-BP model | |
| CN106706314A (en) | Automobile automatic transmission fault diagnosis tester based on virtual instrument and automobile automatic transmission fault diagnosis method based on virtual instrument | |
| Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
| Yang et al. | Computer vision technology based on sensor data and hybrid deep learning for security detection of blast furnace bearing | |
| Gang et al. | Data driven method for spacecraft fault diagnosis: State of art and challenge | |
| Li et al. | Transformer-based meta learning method for bearing fault identification under multiple small sample conditions | |
| CN114444663A (en) | Water supply pipe network leakage detection and positioning method based on time convolution network | |
| CN116817195A (en) | Heating network leakage identification and positioning method based on D-S evidence |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210616 Address after: 226019 No.205, building 6, Nantong University, No.9, Siyuan Road, Nantong City, Jiangsu Province Patentee after: Center for technology transfer, Nantong University Address before: 226019 Jiangsu city of Nantong province sik Road No. 9 Patentee before: NANTONG University |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20181204 Assignee: Nantong Zhijia Intelligent Equipment Co.,Ltd. Assignor: Center for technology transfer, Nantong University Contract record no.: X2022320000355 Denomination of invention: Fault diagnosis method based on multi-sensor signal analysis to provide accurate diagnosis decision Granted publication date: 20200512 License type: Common License Record date: 20221210 |